Non-small cell lung cancers (NSCLCs) from patients exhibit large differences in sensitivity or resistance to chemotherapy and targeted drugs. We hypothesize that these differences will be reflected in tumor mRNA and protein signatures prior to treatment, and that these signatures can be used to improve the effectiveness of therapy. The eventual goal to develop and use such signatures to determine the best available treatment for that individual. To move towards this goal, however, there is a critical need for preclinical models to develop such signatures and test new therapies. This project proposes to use a large panel of NSCLC cell lines and xenografts to systematically measure preclinical therapy response phenotypes, define associated mRNA and protein biomarker signatures of these responses, and then validate these in other cell lines, xenografts, and patient tumor specimens. We will also identify mRNA and protein biomarkers in patient specimens and test them in the preclinical models, eventually resulting in validated biomarkers for response prediction in patients and validated preclinical models. Specific Aims are: Aim 1) To measure quantitative drug sensitivity/resistance phenotypes in a large panel (-100) of human NSCLC cell lines and xenografts (~50), including xenografts made directly from patient tumors without intervening culture, and compare in vitro drug response phenotypes with those of orthotopic (lung) xenografts; Aim 2) To identify microarray mRNA and reverse phase protein array (RPPA)-based expression signatures of response to therapeutic agents in NSCLC lines; using these signatures we will predict drug responses in a new test set of NSCLC cell lines and xenografts, and conduct a preclinical trial comparing the approach of standard non-selected therapy (NST) versus signature-selected therapy (SST); Aim 3) To validate these signatures on available NSCLC specimens clinically annotated as to response to standard and targeted agents (~100 frozen and 200 formalin-fixed paraffin embedded specimens). We will also independently develop response signatures directly from patient samples and test these for predictive ability using preclinical models and other patient specimens. Finally, we will test signatures reported by other investigators and integrate the most informative with our own. From all of this we will develop new methods to select the best available therapies for individual patients, establish a preclinical human tumor model system for systematically testing new drugs, and develop signatures to guide their most efficient use. This project has assembled a multidisciplinary team of basic and clinical investigators, has considerable preliminary data, and clinically annotated tumor specimens for study. It makes use of the Pathology, Biostatistics, and Bioinformatics Cores and interacts with multiple other projects in this SPORE as well as nucleating multiple inter-SPORE collaborations.